2.1.

Tissue Samples

Raman spectra were collected from human coronary artery and aorta tissue samples obtained from explanted recipient heart specimens or at autopsy. To minimize the effects of tissue degradation, the samples were collected within 30 min of surgical excision or autopsy, rinsed with phosphate-buffered saline (PBS), snap frozen in liquid nitrogen, and stored at –85°C until use. Immediately prior to data acquisition, samples were mounted in Histoprep (Colorado Histoprep, Fort Collins, Colorado), and 6-μm-thick cryomicrotome sections were cut and mounted on 1-mm-thick MgF₂ flats (Moose Hill Enterprises Incorporated, Sperryville, Virginia) for spectral microscopy. Contiguous tissue sections were mounted on glass slides, lipid extracted (in 50, 75, and 100% ethanol), rehydrated (in 50 and 25% ethanol), and stained with Oil Red O (ORO) (Poly Scientific Research and Development Corporation, Bay Shore, New York). A total of 66 ceroid deposits were examined: 39 deposits in five samples of coronary artery and 27 deposits in six samples of aorta from 11 patients. Previous studies have shown no significant change in the Raman spectra of atherosclerotic plaque pre- and postfreezing (unpublished data). This should also be true of ceroid deposits, which are already end-oxidized and have deposited in the tissue in an insoluble form.

2.2.

Chemical Species

Raman spectra were acquired from known chemical constituents of atherosclerotic plaque as well as several chemical species postulated in the literature to comprise ceroid deposits. Spectra of cholesteryl linoleate, cholesteryl oleate, cholesterol, collagen, β-carotene, and calcium hydroxyapatite were obtained as previously described.²⁹ The remaining spectra were acquired from chemicals purchased from Sigma (Saint Louis, Missouri) unless otherwise stated. These include: high-density lipoprotein (HDL) (human plasma) and LDL (human plasma); Cu²⁺-oxLDL and MPO/H₂O₂/NO₂-oxLDL (courtesy of Hoff); MPO/H₂O₂/HOCl-oxLDL (courtesy of Hazen); diacetyl dityrosine (courtesy of Sayre); the Maillard reaction product synthesized according to the method of McPherson.;³⁰ the antioxidants L-ascorbic acid, α-tocopherol and retinal palmitate; and hemoglobin (human, A_o ferrous).

2.3.

Instrumentation

The fluorescence and Raman spectra of all tissues and chemical species were acquired using the confocal spectral microscopy system shown in Fig. 1 and described previously.³¹ Briefly, 830-nm light from a Ti:sapphire laser (Coherent Innova 90C/Spectra Physics 3900S, Coherent Incorporated, Santa Clara, California) with a typical power of 150 mW was used to excite the Raman scattering. A 63× objective (0.9 NA) focuses the excitation beam and collects the Raman scattered light in a backscattering geometry. The collected light was passed through a confocal pinhole to achieve a spatial resolution of ~(2 μm)³, filtered by a holographic Raman notch filter (Kaiser Optical Systems, Ann Arbor Michigan), launched into a spectrograph (Chromex 250IS/SM spectrograph monochromator, Albuquerque, New Mexico), and detected using a liquid-nitrogen-cooled charge-coupled device detector (Princeton Instruments, Princeton, New Jersey). The spectrograph has a turret that holds gratings for a range of measurements. For the Raman studies, a 600-grooves/mm grating blazed at 1 μm was used along with the 140-μm spectrograph entrance slit setting, providing ~8-cm^–1 spectral resolution.

Fig. 1

Schematic representation of the confocal Raman and fluorescence spectral microscopy instrumentation.

Simply by switching the excitation source, the filters/ beamsplitters, and spectrograph grating, the system can be used for fluorescence spectral microscopy. For fluorescence measurements, 476-nm excitation light with typical power of 5 mW was delivered to the system via fiber optics from an argon ion laser (Coherent Innova 700, Coherent Incorporated, Santa Clara, California). A 600-grooves/mm grating blazed at 500 nm was used for the fluorescence measurements. The switch between Raman and fluorescence can be made in <1 min.

2.4.

Data Processing

Fluorescence data were typically collected for 0.5–5 s. Raman spectra were collected for 60–500 s to ensure a signal-to-noise ratio adequate to observe species present at low concentrations, such as LDL peroxidation, nitration, and glycation products, which were not detected in the previous Raman ceroid study.³² Often the ceroid deposits had to be photobleached for several minutes to reduce the intense background fluorescence before acquiring a Raman spectrum. All spectral data processing was performed in MATLAB 5.31 (MathWorks, Natick, Massachusetts). Both Raman and fluorescence data were corrected for the spectral response of the system using a tungsten light source and were frequency calibrated using the Raman lines of toluene (Raman) or a mercury lamp (fluorescence). Cosmic rays were removed with a derivative filter and the small background from the MgF₂ flat subtracted. Raman data were fit with a fifth-order polynomial, which was subtracted from the spectra (in addition to the small background from the MgF₂) to remove the fluorescence background. Raman spectra were also corrected for acquisition time and laser power.

2.5.

Spectral Modeling

The chemical composition of the ceroid deposits was determined by analyzing each Raman spectrum with a model. The model uses the Raman spectra of the chemical species described before, called basis spectra. Non-negative least-squares fitting yields the contribution (fit coefficient) of each basis spectrum to the Raman spectrum of the ceroid deposit. The fit coefficients are direct reflections of the concentration of chemical species present in the ceroid deposit. Fit coefficient errors associated with our modeling approach, the noise in the data, and the Raman scattering cross section of each component were assessed using an analytical formula described previously.³³ Model components with fit coefficients less than the error were excluded from analysis, and each Raman spectrum was refit using only those model components with significant fit coefficients. Additional errors in the fit coefficients may be present due to variations in the turbidity of the pure chemical species used to generate the basis spectra and the fluorescence background removal.

3.1.

Raman Spectra of Low-Density Lipoprotein and Its Modification Products

Raman spectra of LDL and its modification products are shown in Fig. 2a. These spectra all have features in common, as they all contain LDL. However, there are distinct Raman bands that indicate differences due to peroxidation, nitration, and glycation effects. Figure 2b displays difference spectra, calculated by subtracting one Raman spectrum from another, for several of the more similar spectra, highlighting distinct Raman bands that reflect changes in lipid and protein structure that ensue during specific modification processes.

Fig. 2

(a) Raman spectra of LDL and its modification products represented in this study. From bottom to top, the spectra are native LDL, Cu²⁺-oxLDL (Fenton Reaction), MPO/H₂O₂/HOCl-oxLDL, MPO/H₂O₂/NO₂-oxLDL, diacetyl dityrosine, and Maillard reaction. (b) Difference spectra, calculated by subtracting one Raman spectrum from another, highlight distinct Raman bands that reflect changes in lipid and protein structure ensuing during specific modification processes. From bottom to top, the spectra are (native LDL) – (Cu²⁺-oxLDL), (native LDL) – (MPO/H₂O₂/HOCl-oxLDL), (Cu²⁺-oxLDL) – (MPO/H₂O₂/HOCl-oxLDL), and (MPO/H₂O₂/NO₂-oxLDL) – (diacetyl dityrosine). Arrows denote Raman bands, discussed in the text, at 1440, 1632, 1655, and 1670 cm⁻¹.

As shown in Fig. 2b, the 1655-cm⁻¹ band, due to a C=C stretching mode indicative of unsaturated bonds, decreases in intensity in the spectra of both Cu²⁺-oxLDL (Fenton reaction) and MPO/H₂O₂/HOCl-oxLDL (MPO-hypochlorite pathway) when compared to the spectrum of native LDL.³⁴ This loss of unsaturation, characteristic of lipid oxidation, is also evident in the 1440-cm^–1 region, assigned predominately to a CH₂ scissoring mode. The broad band at 1655 cm⁻¹ in the spectrum of MPO/H₂O₂/HOCl-oxLDL may represent an interplay between the amide-1 vibration, the lipid C=C stretching mode, and other undetermined bands. The appearance of a band at 1670 cm^–1 in the spectrum of Cu²⁺-oxLDL may indicate a shift from cis to trans isomers ensuing during lipid peroxidation.³⁴

The spectrum of Cu²⁺-oxLDL exhibits a unique band at 1632 cm⁻¹, possibly related to a C=C stretching vibration of conjugated trienes³⁴ or C=O stretch produced by ketone formation.³² This band was present in the Raman spectrum of thermally oxidized lipid published by van der Pol., presented in the context of a study that also employed a combination of fluorescence and Raman spectroscopy to characterize ceroid deposits.³² Although their research elegantly demonstrated the approach, products of lipid peroxidation were not detected. This is likely due to differences in the acquisition times employed in the two studies.

3.2.

Ceroid Composition

Ceroid deposits were presumptively identified by comparison of a phase contrast image of an unstained tissue section used for spectroscopy and a lipid-extracted ORO stained contiguous section. A fluorescence spectrum was obtained from each deposit so identified to confirm fluorescence features characteristic of ceroid. Figure 3 shows the data used to identify one ceroid deposit: a phase contrast image [Fig. 3a]; the corresponding lipid-extracted ORO-stained section [Fig. 3b]; and the fluorescence spectrum acquired from the deposit, highlighted by a box in Fig. 3a [Fig. 3c]. A Raman spectrum was collected from each ceroid deposit identified, and the spectrum fit with a spectral model that yielded the contribution of native and modified LDLs and other chemical species to the ceroid deposit. The accuracy of the model can be qualitatively determined by observing the residual, calculated by subtracting the model fit from the Raman spectrum of the deposit. Figure 3d shows the Raman spectrum of the ceroid deposit presented in Fig. 3a. The model fit and residual are also shown. The small size of the residual indicates a good fit. Although the model employed in this analysis was able to account for the majority of the spectral features observed, small amounts of structure were present in the residuals, indicating that chemical species are present in the deposits at low concentrations or with small Raman scattering cross sections that are not contained in our current model.

Fig. 3

(a) Phase contrast image of a ceroid deposit in a coronary artery plaque (size bar 25 μm). (b) White-light image demonstrating a positive ORO stain for ceroid following lipid extraction. (c) fluorescence spectrum acquired from the deposit, and (d) Raman spectrum acquired from the deposit (solid line) with corresponding model fit (open circles) and residual (below).

Histograms showing the average composition of the ceroid deposits examined in aorta and coronary artery plaques are shown in Figs. 4a and 4b, respectively. The most notable finding in our study is that a variety of mechanisms of LDL modification suggested by in vitro studies are confirmed to contribute to ceroid formation in vivo. In accordance with in vitro predictions, LDL appears within the deposits primarily in its peroxidized form (less native LDL than all LDL peroxidation products detected). Large amounts of free cholesterol and cholesteryl esters are also seen in the deposits because they are byproducts of LDL catabolism within macrophages.³ The primary forms of modified LDL detected in both coronary artery and aortic plaques are products of the Fenton reaction (Cu²⁺-oxLDL) and MPO-hypochlorite pathway (MPO/H₂O₂/HOCl- oxLDL). These two peroxidation products occurred in similar concentrations within the deposits and represented approximately 40 and 30% of the total LDL (native and peroxidized) in the aorta and coronary artery deposits, respectively. Our Raman studies indicate that MPO-mediated LDL modification in ceroid occurs predominately via the hypochlorite rather than the nitric oxide pathway, although diacetyl dityrosine is present in small amounts within the deposits.

Fig. 4

Histograms displaying the average composition of ceroid deposits in (a) aorta and (b) coronary artery plaques. The two standard deviation confidence intervals are shown for each model component. 1. Cu²⁺-oxLDL (Fenton reaction), 2. MPO/H₂O₂/HOCl-oxLDL, 3. MPO/H₂O₂/NO₂ –oxLDL, 4. diacetyl dityrosine, 5. Maillard reaction, 6. cholesterol, 7. cholesteryl linoleate, 8. cholesteryl oleate, 9. HDL, 10. LDL, 11. α-Tocopherol, 12. retinal palmitate, 13. ascorbic acid, 14. β-carotene, 15. collagen, 16. hemoglobin, and 17. calcification.

The presence of collagen, seen in particular in aortic samples [Fig. 4a], is in agreement with ultrastructural studies that have shown ceroid like lipid droplets and protein modification products in association with collagen within atherosclerotic plaques.³⁵ Several antioxidants have been implicated in both the inhibition and promotion of ceroid formation. As shown in Fig. 4, in our Raman studies retinal palmitate and ascorbic acid were detected in ceroid deposits, while α-tocopherol and β-carotene were not.